Bighorn sheep show similar in‐host responses to the same pathogen strain in two contrasting environments

Abstract Ecological context—the biotic and abiotic environment, along with its influence on population mixing dynamics and individual susceptibility—is thought to have major bearing on epidemic outcomes. However, direct comparisons of wildlife disease events in contrasting ecological contexts are often confounded by concurrent differences in host genetics, exposure histories, or pathogen strains. Here, we compare disease dynamics of a Mycoplasma ovipneumoniae spillover event that affected bighorn sheep populations in two contrasting ecological contexts. One event occurred on the herd's home range near the Rio Grande Gorge in New Mexico, while the other occurred in a captive facility at Hardware Ranch in Utah. While data collection regimens varied, general patterns of antibody signal strength and symptom emergence were conserved between the two sites. Symptoms appeared in the captive setting an average of 12.9 days postexposure, average time to seroconversion was 24.9 days, and clinical signs peaked at approximately 36 days postinfection. These patterns were consistent with serological testing and subsequent declines in symptom intensity in the free‐ranging herd. At the captive site, older animals exhibited more severe declines in body condition and loin thickness, higher symptom burdens, and slower antibody response to the pathogen than younger animals. Younger animals were more likely than older animals to clear infection by the time of sampling at both sites. The patterns presented here suggest that environment may not be a major determinant of epidemiological outcomes in the bighorn sheep—M. ovipneumoniae system, elevating the possibility that host‐ or pathogen‐factors may be responsible for observed variation.


| INTRODUC TI ON
Environmental context is thought to shape infectious disease transmission, yet directly comparing wildlife disease events in different environments is difficult due to concomitant differences in host genetics or pathogen strain. One pathogen for which ecological context may be important is Mycoplasma ovipneumoniae, a bacterial agent associated with infectious pneumonia in bighorn sheep (Besser et al., 2008(Besser et al., , 2012(Besser et al., , 2013. M. ovipneumoniae is an important player in the bighorn sheep respiratory disease complex (Dassanayake et al., 2010). Its introduction can lead to severe disease across all age classes of bighorn sheep, generating immediate-term population declines of 10-90% (Cassirer et al., 2018). In the longer term, infected herds may experience years to decades of poor recruitment, driven by sustained infection among a few chronically infected adults Cassirer & Sinclair, 2007;Garwood et al., 2020;Manlove et al., 2016;Plowright et al., 2017). Researchers have speculated that environmental context could have important bearing on the severity of M. ovipneumoniae spillover events (Butler et al., 2018;Manlove et al., 2014), and these speculations are supported by several comparative studies documenting variation in bighorn sheep disease severity across environments (e.g., Butler et al., 2018).
There is evidence that herd mixing patterns may constrain disease burden in lambs (Manlove et al., 2014) and could therefore play a role in herd recovery (Almberg et al., 2021;Dugovich et al., 2017;Lula et al., 2020). Environmental context could also effect epidemiological patterns. For example, varying stress or nutrient availability could produce site-to-site variation in host susceptibility. However, direct comparisons of the same host genetics and pathogen strains in different environments are rare.
Here, we capitalize on intensive sampling surrounding a novel M. ovipneumoniae strain introduction event into members of a single bighorn sheep herd living in two different environments. One group of animals underwent the epidemic in captivity in a holding pen located at Hardware Ranch in northern Utah following capture and translocation from the wild. The other group experienced the epidemic while ranging freely on the herd's original home range near the Rio Grande Gorge in New Mexico. Although the environmental contexts differed, the events stemmed from the same M. ovipneumoniae strain, and animals in both groups had common genetics and health histories. This allowed us to compare disease and symptom progressions, antibody signal strengths, and epidemiological outcomes between the two contexts. Along with the comparisons, we present novel longitudinal data on disease progression during the epidemic, and estimate the incubation period, timing of peak of clinical signs, and patterns of antibody signal strength.

| Study areas and disease event
The Taos Pueblo/Rio Grande Gorge in northern New Mexico (cen-  (1) (Krausman et al., 1985).  During the two captive resampling events, animals were herded into a corner of the pen and chemically immobilized by hand injection with 1.5 ml of butorphanol, azaperone, and medetomidine (BAM, Wildlife Pharmaceuticals, Laramie, WY). After injection, the bighorn sheep were released back into the pen. Once the animals were approachable, they were placed in sternal recumbency, blind folded, and sampled. All animals were reversed simultaneously after sampling with intramuscular injections of naltrexone and atipamezole (Kreeger & Arnemo, 2018). A UDWR veterinarian administered or supervised administration of all drugs.
Nasal and oropharyngeal swabs and serum were collected during the original capture and subsequent animal handling events at Hardware Ranch. Nasal swabs were collected by inserting a single DACRON™ swab into each nostril and gently swabbing the nasal mucosa by swirling the swab. One nasal swab was stored in Tryptic Soy Broth (TSB) for whole-genome sequencing for a different project on the first and second captive sampling events. Diagnostic testing and strain typing were conducted by WADDL. M. ovipneumoniae DNA from six individuals were strain typed using a four-locus MLST method Kamath et al., 2019). WADDL provided cycle threshold values corresponding to PCR diagnostic tests.
Oropharyngeal swabs were gathered during each event except for the March 12th captive sampling, and were submitted to WADDL for bacterial culturing. Blood was collected by jugular venipuncture into PAXGene and serum separator tubes that were centrifuged within 4 h of collection (results presented in Bowen et al., 2022).
Serum was separated and stored frozen in cryogenic vials until it was shipped frozen to WADDL. The sensitivity of the WADDL cELISA test is 90.7%, and its specificity is 95.8%.
At RGG, NMDGF personnel chemically immobilized 29 freeranging animals (19 female animals and 10 male animals) with BAM administered by a dart rifle between April 13th and May 13th, 2020.
Blood and nasal swabs were collected, individuals were assigned a condition score based on palpation, coughing and nasal discharge status were recorded, and animals were fitted with GPS collars (Advanced Telemetry Systems, Isanti MN) for on-going tracking.
Samples were held frozen at NMDGF facilities and then shipped on ice to WADDL for diagnostic testing.
Body conditions at Hardware Ranch were estimated using ultrasound by measuring rump fat and loin thickness of the animals in similar locations as described for deer (Cook et al., 2007) and by palpation. Rump fat, which is the conventional target for ultrasound measurement of ungulate body condition (Cook et al., 2001(Cook et al., , 2007Stephenson et al., 1998), was very low in most animals at the first sampling event (median among positives = 1.25 mm; median among negatives = 1.00 mm). Therefore, we tracked loin thickness as a relative measure of condition changes within an animal in subsequent sampling events.
Five captive animals that died prior to euthanasia were necropsied by pathologists at the Utah Veterinary Diagnostic Laboratory in Logan, UT. The 13 remaining animals were euthanized on March 26th and necropsied by a veterinary pathologist-researcher team.

| Epidemiological outcomes
We tracked survival of captive animals, and although no captive animal delivered a lamb prior euthanasia, we assessed pregnancy status throughout the disease event and weighed and examined fetuses postmortem to estimate fetal ages. Fetal age was estimated based on polished developmental benchmarks for domestic sheep (Sivachelvan et al., 1996) and published birthweights for Rocky Mountain bighorn lambs (Hogg et al., 1992).
The M. ovipneumoniae epidemic at RGG was allowed to play out in the absence of management intervention. Free-ranging animals at RGG were observed by NMDGF biologists over 31 unique observation days between February 25th and July 18th, 2020. Epidemiological data included survival of the instrumented animals captured by NMDGF in April and May; survival of lambs born to instrumented ewes through mid-summer; and aggregate lamb: ewe ratios in the summer and fall (gathered once in July and once in November). Field crews measured lamb: ewe ratios using protocols similar to those used in previous years, which allowed us to compare lamb: ewe ratios in 2020 to prior years.

| Observational scoring of clinical signs and reproductive status
Clinical signs were scored for the captive animals each day.
Observers simultaneously watched all animals in the pen for 45 min and recorded signs including inappetance, nasal discharge, coughing (including number, quality, and pacing of coughs), and lethargy. Nasal discharge at Hardware Ranch was given a numeric score between 0 and 5. Shiny noses were given scores of 1, 2 indicated clear discharge from one nostril, 3 indicated clear discharge from both nostrils, 4 indicated purulent discharge from one nostril, and 5 indicated purulent discharge from both nostrils. Coughs were scored as 1 for isolated coughs, 2 for bouts of five or more consecutive coughs, and 2.5-4 for 2 or more bouts of 5 or more consecutive coughs, depending on depth of cough and number of bouts. Nose licking and head shaking were assigned scores of 1 if present and 2 if consistent throughout the observation period. Total daily scores for each animal were determined by summing the nasal, head shaking, and nose licking scores, and twice the individual's coughing score. Coughing scores were cumulative, with each coughing event contributing separately to the symptom score, so scores had no predefined upper-bound.
Individual daily symptom scores ranged from 0 to 17.5. Photograph records of noses (to track nasal discharge) and hips were gathered as frequently as possible to verify observer records. We did not visually score body condition because animals maintained winter coats throughout the study, but severe declines in body condition were noted on scoring sheets at the observer's discretion.
At RGG, observers recorded coughing (including quality) at the individual level, and nasal discharge status whenever viewers were close enough to see discharge clearly during the summer lamb surveys. However, no observer watched animals at both RGG and Hardware Ranch and scoring rubrics differed between the sites. Therefore, we only compare symptom trends, and not symptom burdens between the sites.

| Study termination and follow-up
UDWR could not identify a suitable location to safely release the captive, infected animals, so the captive animals were euthanized on March 26th, 2020, 34 days after the initial diagnosis. Animals were immobilized with BAM as described above, and sampled live prior to euthanasia via gunshot to the head. Shots were positioned to limit damage to the sinuses, so heads could be assessed for gross evidence of sinus tumors (Fox et al., 2011) at the Utah Veterinary Diagnostic Laboratory. Animals were otherwise necropsied in full at Hardware Ranch by UDWR staff, accompanied by pathologists from UVDL and researchers from Utah State University. Antemortem sampling included collection of the full suite of samples described previously; postmortem sampling included bronchial junction swabs, extraction and measurements of the fetus, and full tissue collection, and gross assessment for sinus tumors.

| Data analysis
2.2.1 | Dynamics of M. ovipneumoniae infection among the captive and free-ranging animals Although WADDL regards PCR results with cycle thresholds above 36 as "indeterminate," we include raw values between 36 and 40 (n = 3) in this analysis. Our goal was to describe how pathogen load changed over the course of infection. As we relied on commercial diagnostic testing, we did not have standard curve values in hand to formally enumerate load at the individual level. However, we were able to track relative load among the samples using the 2^(−Ct') method (Livak & Schmittgen, 2001), using the minimum Ct recorded across all samples at both RGG and Hardware Ranch as the calibrator value. Once sample-specific relative loads were calculated, we compared those relative loads among sampling events and age classes.

| Immune response to M. ovipneumoniae
Serological data analyses aimed to address three main objectives: (1) estimating time to seroconversion among the captive animals; (2) identifying sources of among-individual variation in patterns of antibody signal strength; and (3) comparing serological patterns in the captive and wild settings.
The cELISA test for M. ovipneumoniae should not produce negative "percent inhibition" results, and negative results that are sometimes reported are artifacts of a formula applied to the laboratory-generated data. Therefore, we first reset negative cELISA values to 0, and then added 1 to every reported value to allow for a log transformation. cELISA models were fit using the glm function in the stats package in R, with family set to "Gamma" and an identity link. We compared models with four covariate structures: days post-detection, age, days post-detection plus age, and days postdetection plus linear and quadratic effects for age; assessed model assumptions through standard diagnostic plots; and critiqued model validity by examining predictions arising from each model. Models were compared using AIC, and coefficients were interpreted only for the most parsimonious model within the competitive model set (i.e., the set of models within 2 AIC points of the best-performing model). Dispersion and shape parameter estimates were refined separately after model fitting using the gamma.shape function in R's MASS package (R Core Development Team, 2013), and those estimates were used in the Wald tests to assess coefficient significance.
We estimated the number of days until cELISA percent inhibition exceeded 40 (the WADDL cut-off for classification as non-negative) by fitting a Poisson regression model that treated days since February 22nd as a function of logged cELISA percent inhibition and individual age. We then used that model to predict the day upon which cELISA percent inhibition exceeded 40 across a spectrum of ages.

| Disease progression and severity of clinical signs
Symptom data analyses aimed to address three objectives: (1)  We estimated the time to emergence of clinical signs through a changepoint analysis that described symptom score as a piecewiselinear function of days postexposure using the piecewise. Linear function in R's SiZeR package (Sonderegger, 2012). This analysis identified the most likely breakpoint in the pattern of clinical signs among the captive animals. We built bootstrapped confidence intervals to quantify uncertainty in the changepoint location and pre-and postchangepoint slopes. We used this approach, as opposed to time until first recorded symptoms for each animal to account for occasional sneezes or nasal discharge that are expected among translocated and captive animals independent of M. ovipneumoniae infection.
We described the rise and fall of symptoms postonset by first subsetting the symptom data down to just those data arising after symptom onset (with onset identified using the changepoint identified above), and then fitting a linear model of symptom score as a function of linear and quadratic effects of days postexposure, along with a fixed effect for individual age and a random intercept for each individual.
We compared six different models of body condition (as measured by loin thickness) among the captive animals. All models assumed residual normality and included a random intercept for individual to account for repeated measurements of specific individuals. Fixed effect combinations included models with age; days post-February 22nd; cELISA percent inhibition value; age, days post-February 22nd and an age-by-days interaction term; cELISA, dayspost-February 22nd and a cELISA-by-days interaction term; and a model with cELISA-by-days and age-by-days interaction terms.

| Disease progression and epidemiological outcomes
All captive animals developed symptoms, although the timing and severity of symptoms varied (Figure 3a). Signs were initially mild, with coughing first observed 8 days postexposure. Once symptoms emerged, however, they intensified rapidly, and at the epidemic's peak, we observed over 19 coughing bouts (including one of 85 deep coughs) in a single affected individual-Eartag 35-over a single 45-min observation period. Younger animals had lower clinical sign scores, and clinical signs emerged later in that group (contrast between purple and gold lines in Figure 3a).
Thirteen of the 15 animals that were studied in captivity at Hardware Ranch were censored prior to clearing infection through euthanasia on March 26th ( Table 1) However, lamb survival declined to the range of postdie-off lamb:ewe ratios reported elsewhere (Cassirer et al., 2018). Observers reported symptoms including coughing in lambs on several occasions during summer field observations.

| Dynamics of M. ovipneumoniae infection in the captive and free-ranging herds
Multilocus strain typing Manlove et al., 2019) of M. ovipneumoniae sequences from six different animals indicated that the M. ovipneumoniae strain belonged to a clade TA B L E 1 Longitudinal data from 15 captive female bighorn sheep. "Tag" is eartag number. "Age" is estimated from tooth eruption and wear patterns. "Ct" is the WADDL-derived cycle threshold from a real-time PCR for M. ovipneumoniae (40 corresponds to no detection). "%I" is percent inhibition from the WADDL cELISA serological test (values >40 are regarded as indeterminate, and >50 are regarded as seropositive). "Rump fat" is an ultrasound-based rump fat measurement in mm. "Loin" is an ultrasound-derived measure of loin thickness in mm.
A total of 21 of the RGG animals were PCR-positive for M. ovipneumoniae at capture, and another three were PCR-negative but showed serological signals consistent with exposure (cELISA percent inhibitions of 69, 64, and 64). All three were under four years of age (1, 3, and 3.5), although a formal statistical relationship between age and probability of clearance could not be detected at this sample size. However, the difference in age was detectable in an analysis that pooled the captive and free-ranging animals (p = .048 in a randomization test of animal ages). Among the animals that cleared infection, one was male and three were female.

| Immune response to M. ovipneumoniae
Patterns of increasing antibody signal varied weakly with age among the captive females at Hardware Ranch (Figures 1b and 2a), with seven animals failing to seroconvert by the time of euthanasia (all lines ending at a y-coordinate below 40 in Figure 2a). AIC-based comparisons showed no substantial model improvements when age was included as a predictor (  Table S1).
Antibody expression appeared to follow relatively similar time courses in the captive and free-ranging settings. Although sampling was not concurrent across the two sites, general trends in antibody expression among animals with active M. ovipneumoniae infections aligned well (e.g., Figure 2b). In the free-ranging RGG herd, 11 of 15 animals sampled prior to April 20th had already seroconverted to a cELISA percent-inhibition value above 40 prior to sampling . This difference could be due to random variation in when individuals were sampled relative to their exposure events, which were not well-known at RGG.

| Severity of clinical signs
The changepoint analysis suggested a significant shift in symptoms among the captive animals at 12.9 days postexposure (95% bootstrapped CI [11.50, 18.00]). Symptom scores did not increase significantly with time prior to that day (β days post-exposure pre-changepoint = 0.04; 95% CI

| DISCUSS ION
This study describes in-host dynamics of a novel M. ovipneumoniae introduction into a Rocky Mountain bighorn sheep population, and compares epidemic outcomes in two different ecological contexts.
We estimated a 12.9-day incubation period of very low clinical signs, followed by a period of symptom expansion that was estimated to peak around 36 days postexposure. Seroconversion occurred an average of 24.9 days postexposure among the females held in captivity. Symptom, pathogen load, and serological dynamics varied among individuals, and in one animal, the infection was insufficient to generate an immune response classically indicative of exposure.
Younger animals were more likely to clear the pathogen prior to sampling than older animals. Gross patterns of antibody signal strength and symptom emergence were consistent across the two very different environments, suggesting that pathogen and host-specific factors were important determinants of M. ovipneumoniae epidemiology, at least in this case.

| Comparing disease outcomes in captive and wild settings
The epidemic progressed in very different environments at RGG and Hardware Ranch. While the RGG animals were free-ranging, the captive animals experienced unnaturally high contact rates, along with stress from translocation, living in captivity, dietary shifts, etc. Moreover, data collection at RGG and Hardware Ranch were not perfectly aligned in time, and may reflect different phases of the disease event. While discrepancies in timing may explain some of the differences between the sites, they cannot explain the difference in mortality burdens (18% in captivity vs. 0% in the free-ranging herd). Despite the contextual differences, disease dynamics at the two sites had several similarities. First, antibody signal strengths at Hardware Ranch and RGG followed relatively similar patterns (Figure 2b). The captive data are our only line of information surrounding the period of increasing antibody signal strength during the disease event ( Figure 2a); the RGG data, which arose slightly later in the epidemic, generally showed relatively strong antibody signal strength, especially in the later samples. When temporally aligned with one another, however, the patterns at the two sites are relatively seamless (Figure 2b).
The timing at which clinical signs peaked was also consistent between the two sites: the plateau in clinical signs late in the captive study ( Figure 3a) is consistent with subsequent declining signs in the RGG animals after about April 1st (Figure 3b). The estimated date of maximum symptom burden based on the Hardware data is consistent with the observed maximum symptom burden date at RGG (Figure 3b). However, comparing the intensity of clinical signs is not possible using these data as observation opportunities and overall epidemic intensities may have differed between the captive and wild settings.

| Epidemiology and individual heterogeneity in response to infection
These data provide a unique view of a "goat clade" M. ovipneumoniae strain invading a bighorn herd. Strains derived from domestic goats (as opposed to domestic sheep) cluster genetically (Kamath et al., 2019;Maksimović et al., 2017), and have been shown to produce less-severe infections in experimental settings . In the free-ranging RGG animals, the strain studied here posed a low mortality burden on adult animals relative to .5-year-old and 1.5-year-old animals, respectively, with dashed lines indicating 95% confidence intervals for each fit. (b) Clinical signs (scores are recalibrated to compare data from RGG and hardware ranch; a score of 3 was assigned to animals that were coughing severely at capture, a score of 0 was assigned to animals with no clinical signs). The increase in clinical signs over summer in lamb groups is consistent with endemic-phase M. ovipneumoniae dynamics reported elsewhere . The gray vertical line in B is at 36 days post-February 22nd, the estimated lag until peak symptoms from the captive data what has been reported elsewhere for strains from the "goat clade" , although disease burden on lambs was in line with burdens from more severe M. ovipneumoniae strains (Cassirer et al., 2018;Manlove et al., 2016). Thus, this report suggests a po-  et al., 2021). We cannot determine whether those declines were due to environment or disease in this study. However, loin thickness in the animals that were already infected during the first capture event did not differ significantly (and were typically in incrementally better condition, Figure S2) from loin thickness of uninfected animals during that same sampling event, perhaps providing a weak indication that M. ovipneumoniae alone was not responsible for the body condition declines. Although loin thickness declined with age, younger animals started with lower loin thicknesses, and there may have been a limit in how much loin could be lost.
While all individuals in the captive setting developed infections, symptom severity, infection duration, and antibody responses varied (even within a single host sex, from a single host source population). One proximal driver of this variation was age: younger animals mounted faster antibody responses, had less-severe declines in body condition, and experienced slightly lower pathogen loads.
However, the ultimate processes generating differences among ages are not clear.
Other studies of bighorn sheep pneumonia have reported associations between age and chronic carriage rates , and age and transmission , though the second could not separate behavioral drivers (e.g., young animals had lower contact rates) from in-host factors (e.g., young animals bore lower pathogen burdens or secreted lower volumes of pathogen into the environment). The data presented here suggest that immunodynamics vary with age in the bighorn sheep-M. ovipneumoniae system, consistent with patterns observed in other free-ranging sheep populations (Nussey et al., 2012;Watson et al., 2016).
In the wild setting, we saw differences among sexes in pathogen load and antibody response (both were higher in males). However, our data cannot determine whether those effects are due to differences in date of infection or to differences in immune function between the sexes.

| Study limitations
The data analyzed here arose through a natural experiment, and several factors limit our ability to directly compare individual animals or sites. First, we speculate that most captive animals were exposed on or very shortly after the first capture event on February 22nd, but infection timing is not definitively known. Some animals may have avoided initial exposure, but avoidance would have probably been random (young and old animals were well-mixed in the trailer), and should not substantially confound our analyses.
Timing of exposure in the free-ranging RGG herd is less certain than among the captive animals, thus the symptom dynamics in Figure 3b represent individuals at varying stages of infection.
Nevertheless, the clear decline in symptoms at the RGG herd suggests that most animals resolved acute symptoms of infection within no more than 100 days of exposure. This is fairly consistent with the symptom peak ~36 days postexposure in the captive animals.
Finally, most of data assessed here are from females of reproductive age, and the longitudinal patterns we describe should not be extended to males without validation. A few lines of evidence suggest that symptom dynamics in males may correspond closely to those of females (e.g., the 10 males captured at RGG were not outliers in symptom expression), but we lack sufficient replication to fully document those differences here. formal analysis (lead); investigation (equal); methodology (lead); project administration (lead); resources (supporting); software (lead); supervision (equal); validation (lead); visualization (lead);

ACK N OWLED G M ENTS
We thank Drs. Tom Besser and Frances Cassirer for their consultation on the captive project; and Drs. Logan Weyand, Arnaud Van Wettere, and Jacqueline LaRose for providing necropsy expertise at Hardware Ranch.

FU N D I N G I N FO R M ATI O N
Funding for this project was provided by Utah division of wildlife resources, new Mexico Department of Game and Fish, and Utah State University faculty start-up support for KRM.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available through Dryad at 10.5061/dryad.0vt4b 8h1p.